Farmalytics Project Report

1. Introduction

Soil health plays a crucial role in sustainable agriculture and optimal crop production. This project aims to develop a machine learning-based system to analyze soil conditions and predict two key factors:

• Soil Fertility Classification (Low, Medium, High)
• Irrigation Need Prediction (Irrigation Required or Not)

By leveraging soil parameters such as temperature, humidity, and soil moisture, this system helps farmers make informed decisions to improve crop yield and water conservation.

2. Objectives

• To classify soil fertility based on soil moisture, temperature, and humidity.
• To determine if irrigation is required based on soil moisture levels.
• To develop an automated decision-support system for farmers.
• To optimize water usage and promote sustainable farming practices.

3. Dataset Overview

The dataset used in this project contains three main parameters:

• Temperature (temp): The ambient soil temperature in degrees Celsius.
• Humidity: The percentage of humidity present in the environment.
• Soil Moisture: The moisture content in the soil, which is a key indicator of fertility and irrigation needs.

After preprocessing, two additional columns were added:

• Fertility: Categorized into Low, Medium, and High based on moisture and humidity levels.
• Irrigation Needed: A binary classification (1 = Irrigation Required, 0 = No Irrigation Required).

4. Methodology

Data Collection

We use an ESP32 Dev Kit 1 due to its inbuilt Wi-Fi capabilities. The sensors used are:

• DHT11 for temperature and humidity.
• Capacitive Soil Moisture Sensor 2.0 to collect the moisture data.

This data is sent via HTTP to ThingSpeak, an open-source IoT analytics platform and cloud service that allows users to collect, store, analyze, and visualize real-time sensor data.

Feature Engineering

• Soil fertility was classified using defined moisture and humidity thresholds.
• Irrigation needed was determined based on moisture levels.

Machine Learning Models

• Random Forest Classifier was used to predict soil fertility.
• Decision Tree Classifier was used to predict irrigation requirements.

Evaluation & Optimization

• Model accuracy and classification reports were analyzed.

5. Conclusion & Future Scope

This project demonstrates the potential of machine learning in precision agriculture by providing insights into soil fertility and irrigation needs. Future enhancements could include:

• Incorporating additional parameters like soil pH, nitrogen levels, and rainfall data.
• Implementing real-time sensor data integration for dynamic monitoring.
• Developing a mobile or web-based dashboard for farmers to access insights easily.

With further improvements, this system can significantly contribute to smart farming and sustainable agriculture.